Marketing

A global assessment of the mosaic approach to modeling land surface heterogeneity

Description
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. D14, 4217, /2001JD000588, 2002 A global assessment of the mosaic approach to modeling land surface heterogeneity Andrea Molod 1 Department of Earth
Categories
Published
of 17
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. D14, 4217, /2001JD000588, 2002 A global assessment of the mosaic approach to modeling land surface heterogeneity Andrea Molod 1 Department of Earth and Planetary Sciences, Johns Hopkins University, Baltimore, Maryland, USA Haydee Salmun 2 Department of Geography and Environmental Engineering, Johns Hopkins University, Baltimore, Maryland, USA Received 5 March 2001; revised 20 July 2001; accepted 25 July 2001; published 31 July [1] Modeling the impact of small-scale land surface heterogeneities on scales resolved by general circulation models (GCMs) has long been a challenging problem. We present here a global offline comparison between two approaches to account for the heterogeneities. These approaches are mosaic, which computes separate energy budgets for each surface type within a grid box, and dominant, which assumes that a grid box can be completely described by the dominant vegetation. The experiments are all conducted using the turbulence parameterization of the Goddard Earth Observing System (GEOS) GCM, coupled to the Koster-Suarez Land Surface Model. The results show a large impact in the high- and middle-latitude Northern Hemisphere climates. At high latitudes the warming of the surface after the spring snowmelt is more rapid for dominant. At midlatitudes, where the surface is potentially under moisture stress, the mosaic approach results in a drier, warmer climate. This impact is determined to a large extent by the influence of bare soil areas on the grid-scale climate. The impact of the choice of approach is less important over more homogeneous terrains, such as deserts, as can be expected in the offline framework. These results support the need for a mosaic-type approach to properly model the coupling at the land surface interface. INDEX TERMS: 1878 Hydrology: Water/energy interactions; 3307 Meteorology and Atmospheric Dynamics: Boundary layer processes; 3322 Meteorology and Atmospheric Dynamics: Land/atmosphere interactions; 3337 Meteorology and Atmospheric Dynamics: Numerical modeling and data assimilation; KEYWORDS: general circulation model, mosaic, land surface heterogeneity, hydrology 1. Introduction [2] The energy and material exchanges that occur at the land surface make it a critical component of the Earth s climate system. These exchanges act to partition the net radiation into surface heating, deep-soil heating, and sensible and latent heat fluxes, to redistribute precipitation into evaporation, soil storage, groundwater recharge, and runoff, and to regulate biogeochemical cycles such as photosynthesis, transpiration, the nitrogen cycle, and carbon uptake. The surface fluxes are known to significantly influence rainfall, temperature, and circulation [Milly and Dunne, 1994; Polcher, 1995] from daily to interglacial scales. A recent study shows that the energy exchanges that occur at the land surface are instrumental in effecting the interactions between different modes of climate variability [Wu et al., 1999]. Because so much of the climate signal resides in the 1 Also at Data Assimilation Office, NASA Goddard Space Flight Center, Greenbelt, Maryland, USA. 2 Now at Department of Geography, Hunter College, New York, New York, USA. Copyright 2002 by the American Geophysical Union /02/2001JD000588$09.00 predominant modes of variability other than the mean, capturing these modes and their interactions in global models is crucial to climate prediction [Leetma and Higgins, 1999]. [3] The heterogeneities in the land surface on scales smaller than the typical grid scale of current general circulation models (GCMs) have made it difficult to simulate the mean climate. The difficulty resides in properly capturing the impact of the subgrid-scale variability on the grid scale. Almost all the Soil-Vegetation-Atmosphere- Transfer (SVAT) models that are coupled to state of the art regional and global climate models employ some technique to attempt to account for the subgrid-scale heterogeneities. Comparisons between these different schemes have served to demonstrate the strengths and weaknesses of the schemes and also to aid in the understanding of the influence of small-scale land surface heterogeneity on the atmospheric boundary layer and climate. [4] The earliest of the SVAT formulations for GCMs assumed that the land surface in a GCM grid square can be adequately described by the dominant soil and vegetation characteristics from climatology [Dickinson et al., 1986]. A single set of parameters is specified in the dominant technique which realistically describes the most ACL 9-1 ACL 9-2 MOLOD AND SALMUN: GLOBAL ASSESSMENT OF MOSAIC frequently occurring vegetation and soil type in any GCM grid box. An advantage of this technique (unlike the composite technique described below) is that it specifies only combinations of vegetation characteristics that are found in nature. The dominant technique cannot, however, account in any way for the existence of other vegetation and soil types that may exist over significant areas of the grid box. [5] The majority of the GCMs that are participating in the Atmospheric Model Intercomparison Project (AMIP) II [Gates, 1995] account for the subgrid-scale variability by specifying soil and vegetation parameters that represent a homogeneous composite vegetated surface and its underlying soil for each GCM grid square. Among these are the GCMs used at the European Centre for Medium-Range Weather Forecasts [Viterbo and Beljaars, 1995], the National Center for Atmospheric Research (NCAR) [Gates, 1995], the National Center for Environmental Prediction [Pan and Mahrt, 1987], the Center for Ocean-Land Atmosphere Studies [Xue et al., 1991], the Canadian Climate Centre [Verseghy et al., 1993], and Météo-France [Mahfouf et al., 1995]. There are various different techniques for calculating the appropriate grid-scale vegetation and soil characteristics [e.g., Henderson-Sellers and Pitman, 1992], but in general the parameters are aggregated linearly, except for the roughness length. The aggregated roughness length is computed so as to approximate a linear aggregation of the turbulent momentum flux. A technique that can be viewed as a form of composite is the mixture technique [Sellers et al., 1986], in which up to two vegetation types can occur simultaneously, that is, one atop the other, in a grid box. Mixture is similar to composite in that in both techniques it is assumed that there is horizontally uniform coverage of a combined vegetation type in a grid box. [6] A few of the AMIP II GCMs and some others account for the subgrid-scale heterogeneity using a scheme referred to as the mosaic approach. Separate heat and moisture balance equations are solved for each vegetation type contained within a GCM grid square, and the resulting heat and moisture fluxes which describe the coupling to the atmospheric boundary layer are aggregated linearly. These are the GCMs used at the Goddard Institute for Space Studies [Rosenzweig and Abramopoulos, 1997], Laboratoire Météorologie Dynamique [Ducoudre et al., 1993], the Australian Bureau of Meteorology Research Centre [Desborough and Pitman, 1998], and the NASA/Goddard Seasonal to Interannual Prediction Project [Koster and Suarez, 1992a]. The turbulent diffusion in the boundary layer and above is then computed based on the grid-averaged surface flux of heat and moisture. The statistical-dynamical representations of the subgrid heterogeneity, in which probability distributions for vegetation parameters in a grid box are specified, may be viewed as another type of mosaic representation [e.g., Entekhabi and Eagleson, 1989; Avissar, 1991]. [7] Many intercomparisons exist between the commonly used techniques to account for subgrid-scale variability. Comparisons between dominant and composite, or between different input parameter values in the composite strategy, have been made both on a global and a local scale. Arain et al. [1999] compared the composite and the dominant techniques in 10-year simulations with NCAR s Community Climate Model version 3 and found that the composite technique resulted in potentially large differences in simulated surface fluxes and temperature and improved the simulations over the Sahara desert and the Himalayan mountains. The sensitivity of surface fluxes to the choice of input parameters in a composite technique was examined by Xue et al. [1991]. They showed typical differences in monthly mean latent heat flux of 9 W/m 2 and seasonal differences of 2 W/m 2 when the input leaf area index and albedo were changed based on modern estimates of these parameters. [8] The comparisons between mosaic and composite are mostly local in nature. A direct comparison of the mosaic and composite methodologies during July over the eastern central United States was performed by Klink [1995] and demonstrates an improved simulation of the local climate using the mosaic approach. A comparison of mosaic and mixture for a terrain type which is homogeneously covered by two vegetation types (savannah, where mixture is an accurate representation) was performed by Koster and Suarez [1992b], who showed that the resulting fluxes using the mosaic approach are quite close to the mixture results. Studies by Cooper et al. [1997], van den Hurk and Beljaars [1996], Arola and Lettenmaier [1996], and Polcher et al. [1996] all reported significant differences between mosaic and composite during the spring and summer at several different Northern Hemisphere locations. These studies find that the mosaic representation reduces the latent heat fluxes (and increases the sensible heat) and offers the more accurate estimates of surface fluxes. Mölders et al. [1996] compared the mosaic technique for a central European location in springtime to a fine-mesh model and found that the mosaic technique might tend to underestimate the latent heat flux. [9] The local nature of the comparisons between mosaic and composite makes it difficult to generalize the results to different geographic regions, climate regimes, or different seasons. The present work addresses this issue by focusing on a global comparison of the dominant and mosaic techniques in order to evaluate the impact of modeling technique on the global climate. Differences will be assessed for a varied set of climate regimes and vegetation types throughout the entire annual cycle. This study was conducted in an offline modeling framework, which allows the analysis of direct differences without the additional variability of potential climate feedbacks. The global offline experiments constitute a baseline for fully prognostic studies. The experimental design is described in section 2, along with a brief description of the model used. Results are presented in section 3. Conclusions from this study are presented in the summary in section 4, along with some comparisons with previous studies. 2. Offline Modeling Framework [10] The offline experimental framework is the Offline Land GEOS Assimilation (OLGA), developed by Houser et al. [1997]. OLGA uses the Goddard Earth Observing System-Version 1 Data Assimilation System (GEOS-1 DAS) reanalysis [Schubert et al., 1993] near-surface fields to drive a coupled land surface exchange system. The system consists of the Koster and Suarez [1992a, 1992c] MOLOD AND SALMUN: GLOBAL ASSESSMENT OF MOSAIC ACL 9-3 Figure 1. this issue. A schematic of the OLGA experimental framework. See color version of this figure at back of land surface model (hereinafter referred to as KS LSM) coupled to the GEOS-Terra turbulence and surface layer parameterization [Helfand and Labraga, 1988; Helfand and Schubert, 1995] in the manner described by Molod [1999]. This coupled system uses a mosaic-type approach to model the impact of the subgrid-scale variability with a technique that is called extended mosaic. Extended mosaic differs from a standard implementation of a mosaic scheme in that it extends the independent calculations of the turbulent fluxes over each tile to the top of the model s atmosphere. For the offline experiments of this study, where the atmospheric state is supplied at every time step (as described below), the differences between the standard mosaic and extended mosaic techniques are near zero. We will therefore refer to the extended mosaic experiments as Mosaic (M). OLGA was also used to perform an experiment using the dominant technique (D). [11] A schematic of the OLGA experimental framework is shown in Figure 1. An initial state for the surface and soil is provided from a 5-year GCM simulation to ensure the proper balance between the surface and the deep-soil state. Atmospheric conditions, including temperature, humidity, and winds from the GEOS-DAS (data flow indicated by thin arrows in Figure 1) are used in the GEOS-Terra turbulence scheme to compute turbulent fluxes at the surface and throughout the boundary layer. The sequence of processes in OLGA is indicated by the bold arrows in Figure 1. The turbulent fluxes, along with the net radiation, precipitation, and photosynthetically active radiation from the DAS, are then used by the KS LSM to compute a new surface and soil state. The turbulent fluxes are then updated based on this new surface and soil state to guarantee that the computations conserve energy and moisture. The time integration proceeds with the new state, as indicated by a bold arrow in Figure 1, and with the next input of atmospheric fields from the DAS. [12] The surface and soil state consists of the surface skin temperature, T c (which we call canopy temperature, as is done by Koster and Suarez [1992a]), the deep soil temperature, T d, the near-surface air specific humidity, q a (which we call canopy humidity), the snow amount, the canopy interception reservoir amount, C, and three levels of soil moisture, W shal, W root, and W deep. The temperature budget is solved in each of two layers: the upper one represents the ACL 9-4 MOLOD AND SALMUN: GLOBAL ASSESSMENT OF MOSAIC vegetation canopy and the soil surface, and the lower one represents the deep soil temperature. The moisture budget is solved in each of three layers. The thickness of the top layer ranges from 0.9 to 2 cm and represents surface shallow processes, the second layer represents the root zone and ranges from 0.9 cm to 1.4 m in thickness, and the third layer, referred to as the recharge layer, ranges from 0.3 to 2 m in depth. Above the land surface, OLGA calculates turbulent fluxes of heat and moisture for three atmospheric levels, which approximately constitute the planetary boundary layer. A full boundary layer calculation is performed because turbulence is modeled as a diffusion process with a zero-flux boundary condition at the top of the planetary boundary layer. [13] The OLGA time integration is performed using a time step of 5 min, which is commensurate with the shortest timescales of physical processes at the land surface, and ensures numerical stability of the calculations. Since the fields from the DAS are available at 6-hour intervals, OLGA performs a linear interpolation to obtain forcing data at 5-min intervals. This interpolation scheme has the potential to substantially underestimate the precipitation intensity and thus impact the amounts of runoff and infiltration. These errors may impact the D and M experiments differently. Although more sophisticated algorithms to interpolate precipitation fields have been used (e.g., the Global Energy and Water Experiment Continental-Scale International Project Land Data Assimilation System Project [Mitchell et al., 1999]), the accurate partition of precipitation into infiltration and runoff is not a paramount goal in this study, and the linear interpolation is adequate. [14] The radiative forcing from the DAS in OLGA is the net shortwave and net longwave radiation. Specifying the net radiation implies that two important feedback mechanisms are absent. The albedo is specified implicitly in the net shortwave, based on the snow cover in the DAS, and the snow-albedo feedback, which serves to cool a snow-covered surface, is removed in OLGA. The outgoing longwave is also specified, based on the surface temperature in the DAS, and the ability of a warming surface to be cooled by emitting radiation is also absent. There is, however, a small correction in the outgoing longwave radiation on the order of 1 2% to account for the instantaneous changes in surface temperature. This does not represent a significant longwave radiation feedback. Other offline simulations specify the incoming solar radiation and the downwelling longwave radiation, allowing both the albedo feedback due to the presence of snow and the longwave feedback to respond to changes in surface temperature (The Global Soil Wetness Project (GSWP) [Dirmeyer et al., 1999], for example). This approach provides a more accurate simulation, as was needed for GSWP to obtain accurate estimates of soil moisture. Specifying the net radiation at the surface, however, as is done in OLGA, ensures that the net energy available to the surface is identical in both the D and M experiments. This, in turn, allows us to assume that the causes for the differences between experiments are due to the handling of the surface heterogeneity. [15] A crucial aspect of the present experiments is that an identical model is used in both offline experiments. The dominant and mosaic experiments differ only in their specification of the number of tiles in any grid box. The differences between experiments therefore are solely due to this aspect of handling the land surface heterogeneity and are not due to any differences between models. The intercomparisons between mosaic and composite in other studies, such as those using the Project for Intercomparison of Land-Surface Parameterization Schemes (PILPS) experiments [e.g., Chen et al., 1997], include differences due to model formulation as well as differences in the philosophy of how to handle the heterogeneities. [16] The subgrid-scale variability of the surface is modeled in the KS LSM by viewing each GCM grid cell as a mosaic of independent vegetation stands. The vegetation stands, or tiles, do not interact at all in the OLGA experiments described here. The KS LSM approach to handling subgrid-scale heterogeneities is presented schematically in the mosaic panel of Figure 2, where a hypothetical GCM grid square containing the tiles that describe the mix of surface scene types is shown. Figure 2 also depicts the dominant and composite approaches. In the mosaic approach, all of the bare soil portions of the grid box are treated as though they are juxtaposed, as are all of the deciduous trees, evergreen trees, and shrubs. Each of these types is assigned a fraction of areal coverage, which is used to compute grid-box-averaged fluxes by aggregating linearly. The surface types and the percent of the grid cell occupied by any surface type were derived from the surface classification of Defries and Townshend [1994], and information about the location of permanent ice was obtained from the classifications of Dorman and Sellers [1989]. The geographical distribution of surface designations at 1 1 resolution is shown in Figure 3. Each tile in a grid box, with its own canopy temperature, is assumed to underlie an atmosphere with the same grid mean air temperature. The individual tile values of the change in canopy temperature and humidity computed in the KS LSM are used to provide values of the surface fluxes of heat and moisture, and these surface fluxes are aggregated linearly to provide a single grid-averaged value of the flux across the bottom boundary. [17] We elected to conduct our extended mosaic (M) and dominant (D) experiments using the GEOS-1 DAS reanalysis from 1991 to drive OLGA, because the low Southern Oscillation Index during that year allows us to consider 1991 to be an average climatological period. It should be noted that the GEOS-1 D
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks